Skip to content

monō Foundry

Works with the tools you already use.

Claude Code. Codex. Cursor. VS Code. monō Foundry doesn’t replace your tools, it makes them better: the agentic delivery harness around them, every model, your infrastructure, in production.

Your API keys. Your rates. Zero markup.

$ curl -fsSL https://raw.githubusercontent.com/monoai-labs/mono-foundry/main/install.sh | bash# installs the latest release

$ monofoundry auth login# Google or Microsoft SSO, or --apikey to bring your own

$ monofoundry# your repo, your context, shipping

$▌

Works with Claude Code. Works with Codex. Works with Cursor. Works with VS Code.

Works with

  • Claude Code
  • Codex
  • Cursor
  • VS Code
  • GitHub Copilot

Instant start

Start shipping in under five minutes.

Three commands. No sales call, no slide deck, the first agent runs before the kettle boils.

  1. 01 · Install

    One line in your terminal, macOS, Linux or Windows.

  2. 02 · auth login

    Google or Microsoft SSO, or paste an API key. Your API keys, your rates, zero markup.

  3. 03 · run

    Your repo, your context, shipping. Under five minutes, end to end.

Available on GitHub → Read the CLI before you run it.

The category insight

The harness is the performance variable.

Published benchmarks show the same model jumping from ~2% to 50%+ on SWE-bench from scaffolding alone. The engine matters less than what you mount it in.

Works with your tools

Tools write code. Foundry ships systems.

Your coding agent is the engine. Foundry is the delivery system around it: context engineering, isolated sub-agents, verification gates, typed orchestration, wired into your memory, your connectors, your platform.

Like everything we ship

Sovereign and model-agnostic, by architecture.

Runs in your VPC. Every model, including open-source. The autonomy dial, assist to autonomous, is yours.

The economics

Cheaper than running it raw.

Semantic context chunking, memory-first retrieval and isolated sub-agents mean the same task runs on up to 70% fewer tokens through Foundry than through a coding agent raw. With your own API keys, the saving lands on your own bill.

The Foundry loop

Instruct, orchestrate, verify, ship.

The same loop we ship monō with, your coding agent as the engine, Foundry as the delivery system around it.

  1. 01

    Instruct

    Describe the work in the workspace or your IDE. The project carries its own context.

  2. 02

    Orchestrate

    Foundry plans the work, splits it across isolated sub-agents and mounts your memory and connectors.

  3. 03

    Verify

    Typed gates, evals and human review run before anything lands, nothing ships unchecked.

  4. 04

    Ship

    A pull request on your stack, in your VPC, on the model you chose.

Projects

Start in the workspace. Finish in the IDE.

Draft the PID (project initiation document) with agents in the workspace. Carry the same project into Foundry, the state lives server-side, so nothing is re-explained, re-uploaded or lost.

Start in the workspace

  • PID drafted with agents
  • Work items scoped
  • Context attached

Finish in the IDE

  • Foundry picks up the work items
  • Verification gates pass
  • The pull request ships

One project, server-side: work items · agents · humans · memory

the state never leaves the platform

Agents pick up work items, humans review and redirect, progress visible from both surfaces. From PID to pull request, one project.

Every tile you ship writes to the same memory, your agents don’t start from zero, they start from everything the org already knows. How it compounds →

Client Zero

We ship monō with Foundry.

Client Intelligence, Deal Sourcing and Relationship Intelligence, the products on this site, are Foundry-shipped. The harness you'd buy is the one we bet our own delivery on.

Mid-evaluation? Foundry with your coding agent, honestly compared →

The enterprise path

The sprint is the demo.

Two to four weeks. Our engineers embed with yours and ship something real on your stack. You keep the IP you build on top.

Headless by design

Every primitive, via API.

Orchestration, memory, models, connectors, callable from your own products. Build your own surfaces on the same platform the Collections run on.

API documentation, available in early access.

Your coding agent is only as good as its harness.

Three commands tonight, or a Build Sprint with your team, your call.